2022 Performance for Quantum Computers
After IBM’s Quantum Summit, the new high-performing chip on the scene sets yet another benchmark in what is possible in the world of quantum computing.
Jerry Chow, The Director of Quantum Infrastructure illustrated that the performance of a quantum computer can be broken down to three components.
It is scale, quality and speed. But before I’d focus on these three I’d like to focus on the environment of quantum systems. And for this I interviewed Robert Andrade who was presenting his poster at the 2022 Q Site Conference in Toronto with the title: “Towards Open Quantum Simulations on Quantum Devices”.
Here’s what we discussed:
Question 1 — Patrick Pallagi
In my understanding in an open quantum system we can observe the spin state of a given particle move from one to another.
Does this mean that the more open a quantum system is the less precise we can measure the spins of the particles within?
Answer — Robert Andrade
The main thing about open systems is how vast they are — and how much more massive the possible outcomes of a measurement are. A lot of the time, you want to know whether your energy system has or doesn’t have energy, but as you pointed out, understanding whether it does or does not is a matter of quantum statistical interpretation, and, will a bath that differs very much in temperature from the system, the chance of measuring an up or down spin within the system can become very feeble.
Question 2 — Patrick Pallagi
Can this spin exchange go on indefinitely? Can you tell me more about the constraints for the spin exchange?
Answer — Robert Andrade
From my knowledge, the spin exchange does go on indefinitely, but when we zoom out, the energy exchange between the system and its bath do not go on forever — eventually, we reach equilibrium. However, the idea of equilibrium is not perfect — even in classical mechanics, things that have happened have a probability of occurring again, so our system will always have a probability of snapping back to its initial state during a continuous flow of the spin. The general idea is captured by “Boltzmann recurrence,” or a similarly named idea.
Question 3 — Patrick Pallagi
How do quantum computers tackle the challenge of moving spins, if at all?
Answer — Robert Andrade
Quantum computers are great because they allow us to map mathematically-defined physical entities to qubits. Classical computers do this too, but quantum computers, for my sake, are built in such a way that I can assume my spins exist, and are informed (i.e. there is information in their environment) by the exchange process energies that you mentioned before. We can pretend our qubits are spins in the same way that we can pretend our controlled lasers in the quantum computer are the same as real physical information — both the qubits and spins are physical objects, and they are both constructed to respond to different informations, the laser and the exchange energy, respectively. The information-object pairs are not the same, but there is a logical rule I can apply that transforms one to the other perfectly: thinking of”spin up” as “1”, and “spin down” as “0”.
Now that we have discussed this allow me to focus on the three areas of performance: scale, quality and speed.
What is scale?
Scale is defined by the number of qubits.
We have come a long way from a two qubit system. IBM has been committed to providing a new quantum processor each year.
The highest scale achieved in the past was:
27 qubits in 2019
65 qubits in 2020
127 qubits in 2021
And in 2022, the 433-qubit Osprey chip, unveiled at IBM’s annual Quantum Summit in New York, has more than three times as many qubits as the 127-qubit Eagle chip it introduced last year.
What is quality?
Quality is measured using quantum volume which is a widely adopted benchmark of circuit fidelity that IBM has introduced in 2017. Increasing coherence means better quality.
The question is what is coherence and what is fidelity?
The term ‘quantum coherence’ represents the idea of a superpositioning that is at the heart of quantum mechanics and quantum computing. Specifically, quantum coherence contemplates a situation where an object’s wave property is split in two, and the two waves coherently interfere with each other.
And what is fidelity?
Simply it is the distance (similarity measure) between two quantum states, for example the fidelity between |0⟩ and |1⟩.
To clarify, I’ll try to give some more context.
During the first-ever Pennylane Codecamp organized by Xanadu we were given a coding puzzle that involved noisy circuits.
The puzzle stated:
“There are many ways to combat noise, including a technique called zero-noise extrapolation (ZNE). We will work more with ZNE later! But, on the path towards fault-tolerant quantum computing, we must inevitably understand noise in order to combat it.
PennyLane offers the ability to simulate different types of noise that are present in quantum devices. One such type of noise is called a bitflip error.
A bitflip error, occuring with a probability on a given single qubit, is an error resulting in the state of a qubit being accidentally flipped.
In this coding challenge, you’ll implement a two-qubit circuit that contains bitflip errors on each of the qubits to see how the resulting quantum state drifts from that of the noise-less/error-free circuit. To do so, the fidelity between the respective states for various bitflip probabilities will be calculated.”
So basically, noise can get in the way of having great fidelity, and the more qubits we have the more noisy things can become. That’s why we use noise filtering.
And so Quantum Volume is one of the benchmarks of fidelity.
Infact, quantum volume is a procedure to try and measure the largest effective square circuit you can run.
In Jerry Chow’s presentation he mentions that IBM has managed to create 10^-3 error rates for two qubit gates.
What is speed?
This is a measure of how fast can a system actually solve a problem. IBM is using a metric that they defined as CLOPS ( Circuit Layer Operations Per Second )
The capacity to run a large number of circuits is absolutely critical for targeting quantum advantage as well as applications down the road.
And eventually for error correction, speed is absolutely important looking into the future.
This is because if you can correct errors faster than they can accumulate our fidelity can remain in good standing.
IBM has set themselves a really challenging goal, which was to go from 1400 CLOPS to 10 Thousand CLOPS.
They tackled this by increasing speed in three different ways.
By improving the run-time compiler, the quantum engine and the hardware control systems.
And it has been announced that they have surpassed their goal and achieved 15,700 CLOPS in November of 2022.
Exciting news for the world.
A promising update for the quantum community.
Thank you for reading!